Search papers, labs, and topics across Lattice.
ARCS is introduced as a system for amortized analog circuit generation, combining a graph VAE and flow-matching model with SPICE-based ranking. It addresses a key limitation of REINFORCE by introducing Group Relative Policy Optimization (GRPO), which normalizes advantages per topology to mitigate cross-topology reward distribution mismatch. Results show ARCS achieves 99.9% simulation validity with significantly fewer SPICE evaluations than genetic algorithms, enabling rapid prototyping and design-space exploration despite not yet matching the per-design quality of search-based optimization.
Analog circuit designs can now be generated 1000x faster thanks to a novel approach that combines learned generators with SPICE-based ranking and a fix for REINFORCE's cross-topology reward distribution mismatch.
I present ARCS, a system for amortized analog circuit generation that produces complete, SPICE-simulatable designs (topology and component values) in milliseconds rather than the minutes required by search-based methods. A hybrid pipeline combining two learned generators (a graph VAE and a flow-matching model) with SPICE-based ranking achieves 99.9% simulation validity (reward 6.43/8.0) across 32 topologies using only 8 SPICE evaluations, 40x fewer than genetic algorithms. For single-model inference, a topology-aware Graph Transformer with Best-of-3 candidate selection reaches 85% simulation validity in 97ms, over 600x faster than random search. The key technical contribution is Group Relative Policy Optimization (GRPO): I identify a critical failure mode of REINFORCE (cross-topology reward distribution mismatch) and resolve it with per-topology advantage normalization, improving simulation validity by +9.6pp over REINFORCE in only 500 RL steps (10x fewer). Grammar-constrained decoding additionally guarantees 100% structural validity by construction via topology-aware token masking. ARCS does not yet match the per-design quality of search-based optimization (5.48 vs. 7.48 reward), but its>1000x speed advantage enables rapid prototyping, design-space exploration, and warm-starting search methods (recovering 96.6% of GA quality with 49% fewer simulations).